If the two tests were applied to two independent samples, use the Chi-square test. If applied on the same sample, use the McNemar chi-square.

thanks for kind response dear

for example: grace score  and TIMI score applied to the same patients to predict major adverse cardiac events. herein, which test??

(Hypothesis and significance-) tests are about data related to a stocastic model, not about properties of diagnostic scores.

Dear sir,

i am attached two articles you have read it.

https://www.ncbi.nlm.nih.gov/pubmed/3203132

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2637106/

http://www.spl.harvard.edu/archive/spl-pre2007/pages/papers/zou/acad-radiol01/reproc/reproc.pdf

I am not sure what the practical meaning of the NRI is (as in [email protected]).

Hi Mohamed Salah Abd ELbasi,

I suggest, convert the two scores to standardized z-scores, use logistic regression to compute OR (95% CI) for predicting major adverse cardiac events by the z-scores of the two test, and look at the AUCs. Attached please find my paper.

BW

Riaz

 Further to my suggestion, if you looking at the agreement between the two scores, you may look at the Bland Altman analysis for agreement, provided the test scores are measured on the same scale. if you are interested in looking at the difference between the two scores, then use paired-value t-test. 

Dear all

Thank you for kind response

my outcome in major adverse cardiac events is yes or no (not percent) therefore, I think that net reclassification tables (NRT) will not help in that situation

I think that odds ratio calculation is logic in that situation

But my question to Dr Muhammad Riaz about applying paired value t test (is it applied on the probability of each value of the two scores that are automatically calculated by binary logistic regression in spss?) 

Dear Mohamed Salah Abd ELbasit,

As your outcome is binary and you are assessing the predictive ability of  the two type of scores (I don't know what the scores are though), better analysis this is: 

convert the two scores to standardized z-scores, use logistic regression to compute OR (95% CI) for predicting major adverse cardiac events by the z-scores of the two test, and look at the Area under the ROC curve and the ORs to decide which of the two scores is better in predicting  outcome.

Dear Dr Muhammad Riaz

Thank you for elegant share

but can I calculate a p value of the difference between two scores?

Greetings

If  you  are looking at the mean difference between the two scores, I suggest use paired ttest

sincere thanks Dr Mohammad for your

kind response